Energy data synthesis

Abstract

A system and corresponding method for synthesizing energy time series data for a facility is provided. The system includes a time series generator (100) configured to receive an input set of attributes (102) comprising attributes characterizing the facility and to output synthesized time series (104) representing estimated energy data for the facility. The synthesized time series is generated based on the input set of attributes (102) and one or more reference time series (108) associated with a respective reference set (106) of attributes.

Description

Technical Field

[0001] The present invention relates to the synthesis of energy time series data for facilities such as industrial sites.

Background Art

[0002] In order to optimize the electrical energy consumption of an industrial site, it may be useful to have access to detailed time series data giving regular values (e.g. every 15 minutes) of consumption, cost, and / or environmental impact over a particular time period (e.g. one year). However, access to real data may not be available for various reasons, such as having to estimate the time series data. For example, a general or "typical" time series may be selected from a database and considered to be sufficiently similar to the expected consumption profile. However, such typical time series are often unavailable. Real consumption data covering a limited time period (e.g. one day or one week) may be obtained and repeated to fill a full year, but doing so typically eliminates the influence of seasonal trends.

Summary of the Invention

[0003] Therefore, there is a need to increase the accuracy of time series estimates without accessing the real time series data of the facility under investigation. This need is met by the subject matter of the independent claims. Optional features are described by the dependent claims.

[0004] According to a first aspect, a system for synthesizing energy time series data for a facility is provided. The system comprises an input set of attributes comprising attributes characterizing the facility, and a time series generator configured to output a synthesized time series representing the estimated energy data for the facility. The synthesized time series may be generated based on the input set of attributes and one or more reference time series associated with respective reference sets of the attributes.

[0005] The claimed subject matter enables the generation of a time series that is valid for a specific facility requiring only a limited set of key attributes, using a set of exemplary facilities for which key attributes and time series are known. The constructed time series may have sufficient granularity and time span to provide reliability for the results (i.e., the predicted optimization potential). The time series can be used directly by the optimizer, and additional (e.g., manual) steps may not be required. The claimed system and method can be easily used by non-technical experts. The generated time series can be used for the design and development of energy management software covering a wider range of examples. Customers do not need to provide detailed data. The claimed subject matter thus facilitates the estimation of energy cost optimization for commercial or industrial facilities or commercial or industrial areas.

[0006] Time series generators can synthesize time series by directly modifying existing (e.g., similar) time series, or by synthesizing them from scratch, for example, using machine learning models. Time series generators may also be called time series generation modules.

[0007] For example, a time series generator may include a time series modification module configured to produce a composite time series by modifying one or more reference time series, where the reference set of attributes is determined to be similar to the input set of attributes. The time series modification module may also be called a fitting module.

[0008] To facilitate the identification of appropriate time series for correction, the time series generator may include a reference data retrieval module configured to search one or more databases to retrieve one or more candidate time series associated with each candidate set of attributes, and to select one or more candidate time series associated with attribute candidate sets similar to the input set of attributes as reference time series. The retrieval module may also be called a search and compare module or a database retrieval module.

[0009] Any appropriate determination of similarity may be used. For example, the reference data acquisition module may be configured to select as one or more reference time series associated with a set of attribute candidates that, when compared to an input set of attributes, yield a similarity metric that satisfies a given similarity threshold. The similarity metric may be expressed as a distance metric. Thus, in this example, the reference data acquisition module may be configured to determine the similarity metric for each set of attribute candidates by calculating a distance metric that represents the distance between the set of attribute candidates and the input set of attributes. In addition or alternatively, in this example, the reference data acquisition module may be configured to determine the similarity metric for each set of attribute candidates by inputting two such sets of attributes as feature vectors into a trained classifier trained to predict a similarity metric targeting an input feature vector comprising the set of attribute candidates and the input set of attributes.

[0010] Various methods for modifying a reference time series are envisioned in this disclosure. In one example where one or more reference time series comprise multiple reference time series, the time series modification module may be configured to modify the reference time series by calculating a weighted combination of the multiple reference time series and outputting the weighted combination as a composite time series. The weighted combination may be determined or calculated, for example, as cTS = 1 / D * Σ [di * eTSi], where the reference set of attributes eSAi is similar to the input set of attributes cSA within a distance of di, and D = Σdi. The time series modification module may be further configured to use an optimization algorithm to find the optimal weights for the weighted combination. The weights may represent the customer time series, for example, as a linear combination of the multiple reference time series. In an alternative form of modification, the time series modification module may be configured to modify multiple reference time series by concatenating selected intervals from the multiple reference time series, where the intervals are selected according to the similarity between the reference set of attributes associated with each of the multiple reference time series and the input set of attributes.

[0011] Instead of modifying existing time series, or even modifying them, a time series generator may be configured to generate a synthesized time series by inputting a set of attributes into a machine learning model that has been trained to predict one or more reference time series based on each of the reference sets of attributes input into the machine learning model. One or more reference time series can serve as targets that are predicted based on each of the reference sets of attributes input into the model as feature vectors.

[0012] Various post-processing steps may be performed to refine the synthesized time series. For example, the time series generator may be further configured to modify the synthesized time series in the post-processing steps to match one or more predetermined important attributes of the facility. For example, the generated energy profile may be rescaled to match the facility's total energy consumption or generation, or peak and average power. Another possible post-processing step involves adjusting the generated time series to ensure that the load pattern follows the production pattern. For example, if the facility has an industrial site, and it is priori known that production at the site is not running during certain time periods (e.g., nighttime, off-peak hours, or weekends), the generated time series (load profile) may be adjusted by point-by-point multiplication with a time series that is, for example, 1.0 at points during full production and, for example, 0.0 or a small positive value (below a predetermined threshold) at points during off-peak hours. Linear interpolation or another suitable interpolation approach may be used to determine the values ​​at points during the transition between full production and no / limited production.

[0013] According to a second embodiment, a method is provided for synthesizing energy time series data for a facility. The method may comprise receiving an input set of attributes having attributes that characterize a facility, and outputting a synthesized time series representing estimated energy data for the facility. The synthesized time series may be generated based on the input set of attributes and one or more reference time series associated with each reference set of attributes.

[0014] According to a third aspect, a computer program product is provided which, when executed by a computer, comprises instructions that enable the computer to carry out the method described in the second aspect.

[0015] According to a fourth aspect, a computer-readable medium is provided, which, when executed by a computer, enables the computer to carry out the method described in the second aspect.

[0016] In another embodiment, a method and / or system for optimizing energy consumption in a facility based on a time series synthesized as described herein is provided. In this regard, known energy consumption optimization methods may be used.

[0017] As used herein, the term “module” may be replaced by “system,” “circuit,” or “tool,” depending on the specific implementation.

[0018] "Attributes" means any characteristics or parameters that can characterize or specify a facility, and in particular, factors that relate to or influence the energy consumption or generation of the facility. For example, attributes may include any one or more of the following: the classification of the facility, the annual energy consumption or generation at the facility, operating hours, geographical location, and the equipment used by the facility.

[0019] "Facility" means a portion of land, buildings, or equipment that consumes and / or generates energy. Therefore, a facility may be described as an energy consumer, an energy generator, or both. A facility may be a commercial or private facility, such as an office building or a home, or an industrial facility, particularly an industrial site such as an industrial plant, including any combination of power generators, consumers, and storage equipment. The term "facility" encompasses not only a portion of land, buildings, or equipment as a whole, but also their individual parts, sections, or components.

[0020] The terms "generate" and "combine," as used in relation to time series, can be replaced by equivalents such as "construct," "build," "make," and "create."

[0021] The adjective “reference” in the terms “reference time series” and “reference set of attributes” as used herein may be replaced with equivalents such as “example,” “existing,” “pre-stored,” or “template.” The term “input set of attributes” is also discussed herein in relation to “customer site attributes.”

[0022] As used herein, the term “candidate” refers, in particular, to a set of time series or attributes that are being compared to or selected from others.

[0023] The time series in the following examples relate to energy consumption data, but the time series may also relate to energy generation data, either in addition or alternatively. Thus, “energy data” as it appears herein can be understood to include energy consumption data, energy generation data, or both. Equivalent terms such as energy exchange data or energy transfer data may be used. Furthermore, while the specific examples given below primarily relate to electrical energy time series, it will be understood that the systems and methods described are equally applicable to other time series, such as thermal energy time series, or to time series relating to other entities (such as price, weather, or production volume).

[0024] The present invention includes one or more aspects, embodiments, or features, either individually or in various combinations, whether expressed in combination or individually (including in the claims). Any optional feature or sub-aspect of one of the above aspects may be applied to any of the other aspects as appropriate.

[0025] The above embodiments will become clear from the following embodiments for carrying out the invention and will be explained by reference thereto.

[0026] Here, the embodiments for carrying out the invention are given by way of example only, with reference to the accompanying drawings.

Brief Description of the Drawings

[0027] [Figure 1] An example of a time series generator is illustrated. [Figure 2] An implementation form of the time series generator in FIG. 1 is illustrated. [Figure 3] A further implementation form of the time series generator in FIG. 1 is illustrated. [Figure 4A] Various illustrative time series are illustrated. [Figure 4B] Various illustrative time series are illustrated. [Figure 4C] Various illustrative time series are illustrated. [Figure 4D] Various illustrative time series are illustrated. [Figure 5A] The use of a neural network in generating a time series is illustrated. [Figure 5B] The use of a neural network in generating a time series is illustrated. <000O112> [Figure 6] A computing device that can be used in accordance with the systems and methods disclosed herein is illustrated. <000011S>

Embodiments for Carrying Out the Invention

[0028] FIG. 1 illustrates a time series generator 100 configured to generate an energy time series from important attributes of a site. The time series generator 100 receives, as input, a customer site attribute (cSA) 102 that includes important attributes of the site in question. The time series generator 100 is configured to generate or synthesize a customer time series (cTS) 104 for the site based on the customer site attribute 102 in conjunction with an existing site attribute (eSA) 106 and respective existing time series (eTS) 108 provided by one or more databases.

[0029] Examples of site attributes 102 and 106 include one or more of the following: site classification (e.g., individual manufacturing, commercial retail, food production, etc.), annual site energy consumption (e.g., 5 GWh), site operating hours (e.g., given by the number of working days per week and day shifts), site geographical location (e.g., country or GPS coordinates), and equipment used by the site (e.g., batteries, solar panels, etc.).

[0030] Figure 2 illustrates one implementation of the time series generator of Figure 1, in which the time series generator 100 is configured to synthesize time series by directly modifying one or more existing similar time series 108. In other words, the time series generator 100 is configured to generate a customer time series 104 using a mapping from site attributes to time series. In one example, exemplary site attributes 106 are clustered by, for example, k nearest neighbors, principal component analysis for dimensionality reduction (e.g., taking only the five components that give the maximum variance contribution), and subsequent k nearest neighbors. The exemplary time series 108 are also clustered by time series clustering algorithms such as k nearest neighbors, multivariate principal component analysis, or by using the absolute value of the Pearson correlation coefficient if the absolute value of the Pearson correlation coefficient exceeds a predetermined threshold (e.g., |correlation|>=0.5). Other suitable clustering methods are also assumed by this disclosure. To generate customer time series 104, a land attribute cluster corresponding to customer land attribute 102 is determined (for example, based on similarity as described below), and the exemplary time series 108 selected as candidate modifications for generating customer time series 104 include a first set of exemplary time series 108 associated with exemplary land attribute 106 in the determined land attribute cluster, and optionally a second set of exemplary time series 108 found in the same time series cluster(s) as those in the first set.

[0031] The time series generator 100 includes a reference data acquisition module 202 configured to search a database of existing land attributes 106 and identify entries that are most similar to customer land attributes 102 (which can be clustered in the manner described above). The reference data acquisition module 202 is configured to calculate the similarity between customer land attributes 102 and one or more candidate sets of exemplary land attributes 106 in the database in one or more ways of: a) by defining an appropriate distance metric that measures the distance between two sets of attributes (represented as vectors in a multidimensional space); or b) by using a machine learning classification method.

[0032] The time series generator 100 further comprises a time series modification module 204 configured to construct one or more corresponding customer time series 104 using an identified set of similar exemplary land attributes 106. Assuming that a set of exemplary land attributes eSAi 106 is similar to customer land attributes cSA 102 within a distance of 1 / di, where D = Σdi, the customer time series 104 can be constructed using, for example, one or more of the following methods:

[0033] Referring to Figures 4A-C, the first method calculates a weighted combination of relevant exemplary time series, e.g., cTS = 1 / D * Σ [di * eTSi]. Figure 4A shows an example of an exemplary time series 108 for a site where the load has a day / night pattern. Figure 4B shows a further example of an exemplary time series 108 for a site where the load is switched on and off periodically (more frequently than the frequency of the diurnal cycle shown in Figure 4A). Figure 4C shows a weighted combination of the time series from Figures 4A and 4B using weights d1=3, d2=1.

[0034] A typical post-processing step might involve formulating the optimization problem to find the best-fitting weights that represent the customer time series 104 as a linear combination of example time series 108, where several important properties must be satisfied. The optimization problem may eliminate the need to perform post-processing steps to ensure that certain properties are satisfied, which are rather included as constraints in the optimization problem. For example, instead of using fixed weights (e.g., di) in the linear combination of example time series, the optimization problem can be formulated as follows: Minimize ||(t1-d1, t2-d2, ..., tn-dn)||_p^p under the conditions t1>=0, ..., tn>=0, cTS=Σti*eTSi, total energy consumption of cTS=given total energy consumption, while satisfying any further major properties. The optimization variables are given by t1, ..., tn, where || ||_p represents the p-norm of the vector, where typically p=2 or p=1. In other words, this optimization problem means finding the set of coefficients ti that are closest to di in the sense of the p-norm, such that the positive linear combination cTS = Σti * eTSi matches the total energy consumption of the customer's site.

[0035] The second method involves selecting only a predetermined number of the most similar plots (as discussed above, according to the similarity of plot attributes) and concatenating different intervals from various exemplary time series 108 associated with the selected plots. Here, rather than using a linear combination of exemplary time series 108, various exemplary time series 108 are used to generate customer time series 104 by concatenating selected intervals of different exemplary time series 108. The similarity metric described above can be used to determine the relative proportion of the different intervals selected from the various exemplary time series 108. For example, the weight di can be used to select the proportion of time intervals. An example is shown in Figure 4D: The similarity of the plot attributes under investigation is high in Example A (Figure 4A) (d A =3), in example B (Figure 4B) it is moderate (d BIt was found that = 1). This is a combination of 3 days for example site A and 1 day for example site B, which is then randomly placed as day 2 in Figure 4D, resulting in the resulting time series (Figure 4D).

[0036] In the post-processing step, the constructed data can be made more accurate by precisely matching it to key characteristics of the customer's site. For example, the generated load profile can be rescaled to more closely match the site's total energy consumption.

[0037] Figure 3 illustrates a further implementation of the time series generator in Figure 1, in which the time series generator 100 features a trained model 304 (e.g., a neural network or a recurrent neural network) trained using a sufficient number of exemplary land attributes and corresponding exemplary time series from a database, with the goal of constructing a time series similar to an exemplary time series based on a given set of land attributes. The trained model 304 is used to generate a customer time series 104 from specified customer land attributes 102. The time series generator 302 also optionally includes a model training module 302 configured to train the model 304 based on exemplary land attributes 106 and an exemplary time series 108. Alternatively, a pre-provided model 304 may be used.

[0038] Figures 5A and 5B illustrate an illustrative topology of such a recurrent neural network (RNN). The RNN predicts the next value of the customer time series 104, taking into account the current value of the time series and the customer site attribute 102. The first value of the customer time series 104 can be taken as a random number within a suitable boundary. For example, a suitable boundary for a load profile for a small industrial site might have a lower bound of 0 MW and an upper bound of 1 MW. The boundary may be included as part of the site attribute or may be inferred from the site attribute. Training of such a neural network can be performed by learning the example site attribute 106 and the example time series 108 as inputs to a shifted example time series 108 as an output, where "shifted" means that all values ​​are shifted one time step into the future. In addition, the example input time series 108 may be truncated so that the last time point is removed. Training can be performed using one or more of the following illustrative methods: mean squared error (MSE) as the loss function; L2 regularization of neural network weights; any suitable variation of a stochastic gradient descent algorithm such as the Adaptive Moment Estimation Algorithm (Adam) for loss minimization; and a learning rate parameter (typical values ​​are 10). -5 ~10 -1 (within the range), L2 regularization parameter (typical value is 10) -8 ~10 -1 Hyperoptimization by random sampling of parameters that define the number of hidden layers and the size of units in the network layer (typical values ​​are shown in Figure 5B, but can be varied within a range of, for example, 10 times smaller / larger, and can be powers of 2).

[0039] In a modified form, a generative adversarial network may be used to generate customer time series 104 using machine learning methods.

[0040] It will be understood that the above disclosures provided in relation to commercial / industrial areas are not limited in this respect and may be applicable to the generation of any energy time series.

[0041] Referring here to Figure 6, an illustrative high-level diagram of an exemplary computing device 800 that can be used in accordance with the systems and methods disclosed herein. The computing device 800 includes at least one processor 802 that executes instructions stored in memory 804. Instructions may be, for example, instructions for performing a function described above as being performed by one or more components discussed above, or instructions for performing one or more of the methods described above. The processor 802 may access memory 804 via a system bus 806. In addition to storing executable instructions, memory 804 may also store conversation inputs, scores assigned to conversation inputs, etc.

[0042] The computing device 800 also includes a data store 808 accessible by the processor 802 via the system bus 806. The data store 808 may contain executable instructions, log data, etc. The computing device 800 also includes an input interface 810 that allows external devices to communicate with the computing device 800. For example, the input interface 810 may be used to receive instructions from an external computer device, a user, etc. The computing device 800 also includes an output interface 812 that interfaces the computing device 800 with one or more external devices. For example, the computing device 800 may display text, images, etc. via the output interface 812.

[0043] External devices communicating with the computing device 800 via the input interface 810 and output interface 812 are intended to be included in an environment that provides substantially any type of user interface with which the user can interact. Examples of user interface types include graphical user interfaces and natural user interfaces. For example, a graphical user interface may accept input from a user using an input device such as a keyboard, mouse, or remote control, and provide output on an output device such as a display. Furthermore, a natural user interface may allow the user to interact with the computing device 800 in a way that is free from the constraints imposed by input devices such as keyboards, mice, and remote controls. Rather, a natural user interface may rely on speech recognition, touch and stylus recognition, gesture recognition both on and adjacent to the screen, air gestures, head and eye tracking, voice and speech, vision, touch, gestures, machine intelligence, etc.

[0044] Furthermore, although it is illustrated as a single system, it should be understood that computing device 800 can be a distributed system. For example, several devices may communicate via network connections and collectively perform tasks described as being performed by computing device 800.

[0045] The various functions described herein can be implemented in hardware, software, or any combination thereof. When implemented in software, these functions can be stored on or transmitted through a computer-readable medium as one or more instructions or codes. Computer-readable medium includes computer-readable storage media. Computer-readable storage media can be any available storage medium accessible to a computer. Such computer-readable storage media may include, but are not limited to, RAM, ROM, EEPROM®, CD-ROM or other optical disk storage devices, magnetic disk storage devices or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that is accessible to a computer. Disk and disc, as used herein, include compact disc (CD), laserdisc (disc), optical disc (disc), digital versatile disc (disc) (DVD), floppy disk (disk), and Blu-ray disc (disc) (BD), where a disk typically reproduces data magnetically and a disc typically reproduces data optically using a laser. Furthermore, propagated signals are not included within the scope of computer-readable storage media. Computer-readable media also include communication media, which include any medium that facilitates the transfer of computer programs from one place to another. A connection may be, for example, a communication medium. For example, if software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of a communication medium. Any combination of the above should also be included within the scope of a computer-readable medium.

[0046] Alternatively, or in addition, the functions described herein can be performed, at least in part, by one or more hardware logic components. For example, but not limited to, exemplary types of hardware logic components that can be used include field-programmable gate arrays (FPGAs), program-specific integrated circuits (ASICs), program-specific standard products (ASSPs), system-on-chip systems (SOCs), composite programmable logic devices (CPLDs), and the like.

[0047] The applicant hereby discloses each individual feature described herein and any combination of two or more such features, individually, to the extent that such features or combinations can be implemented as a whole in light of the common general knowledge of those skilled in the art, regardless of whether such features or combinations of features solve any problem disclosed herein, and without limiting the scope of the claims. The applicant shows that aspects of the present invention may consist of any individual features or combinations of features.

[0048] The present invention has been illustrated and described in detail in the drawings and the foregoing description, but such illustrations and descriptions should be considered illustrative and not limiting. The present invention is not limited to the disclosed embodiments. In consideration of the foregoing description and drawings, it will be apparent to those skilled in the art that various modifications can be made within the scope of the present invention as defined by the following claims. The following is a direct reproduction of the claims as originally filed. [1] A system for synthesizing energy time-series data for a facility, wherein the system is A system comprising a time series generator (100) configured to receive an input set (102) of attributes having attributes that characterize the facility, and to output a synthesized time series (104) representing estimated energy data for the facility, wherein the synthesized time series is generated based on the input set (102) of attributes and one or more reference time series (108) associated with each reference set (106) of attributes. [2] The system according to [1], wherein the time series generator (100) comprises a time series modification module (204) configured to produce the combined time series (104) by modifying one or more reference time series, and the reference set of attributes is determined to be similar to the input set of attributes. [3] The system according to [1], wherein the time series generator (100) includes a reference data acquisition module (202) configured to search one or more databases to retrieve one or more candidate time series associated with each candidate set of attributes, and to select one or more reference time series (108) associated with candidate sets of attributes that are similar to the input set of attributes (102). [4] The system according to [3], wherein the reference data acquisition module (202) is configured to select one or more reference time series (108) associated with a candidate set of attributes that yield a similarity metric that satisfies a predetermined similarity threshold when compared with the input set of attributes (102). [5] The system according to [4], wherein the reference data acquisition module (202) is configured to determine the similarity metric for each candidate set of attributes by calculating a distance metric representing the distance between the candidate set of attributes and the input set of attributes (102). [6] The system according to [4], wherein the reference data acquisition module (202) is configured to determine the similarity metric for each candidate set of attributes by inputting two such sets of attributes as feature vectors to a trained classifier trained to predict a similarity metric target based on an input feature vector comprising a candidate set of attributes and an input set (102) of attributes. [7] The system according to any one of [2] to [6], wherein the one or more reference time series comprises a plurality of reference time series (108), and the time series modification module (204) is configured to modify the reference time series by calculating a weighted combination of the plurality of reference time series and to output the weighted combination as the combined time series (104). [8] The time series modification module (204) is further configured to use an optimization algorithm for finding the optimal weights for the weighted combination, as described in [7]. [9] The system according to any one of [2] to [6], wherein the one or more reference time series comprises a plurality of reference time series (108), and the time series modification module (204) is configured to modify the plurality of reference time series by concatenating intervals selected from the plurality of reference time series, the intervals being selected according to the similarity between a reference set of attributes associated with each of the plurality of reference time series and an input set of attributes.

[10] The system according to [1], wherein the time series generator (100) is configured to generate the synthesized time series (104) by inputting an input set of attributes (102) into the machine learning model (304) which has been trained to predict one or more reference time series (108) based on each reference set (106) of the attributes input to the machine learning model (304).

[11] The system according to any one of [1] to

[10] , wherein the time series generator (100) is further configured to modify the synthesized time series (104) in a post-processing step to match one or more predetermined key attributes of the facility.

[12] The system described in any one of [1] to

[11] , wherein the attributes include one or more of the following: the classification of the facility, the annual energy consumption or generation at the facility, the operating hours, the geographical location, and the equipment used by the facility.

[13] A method for synthesizing energy time series data for a facility, the method being The process involves receiving an input set (102) of attributes that characterize the aforementioned facility, Output a composite time series (104) representing the estimated energy data for the aforementioned facility. A method comprising, wherein the synthesized time series is generated based on an input set of attributes (102) and one or more reference time series (108) associated with each reference set (106) of the attributes.

[14] A computer program product comprising instructions that, when executed by a computer, enable the computer to carry out the method described in

[13] .

[15] A computer-readable medium comprising instructions that, when executed by a computer, enable the computer to carry out the method described in

[13] .

Claims

1. A system for synthesizing energy time-series data for a facility, wherein the system is The system includes a time series generator (100) configured to receive an attribute input set (102) having attributes that characterize the facility, and to output a synthesized time series (104) representing estimated energy data for the facility, wherein the synthesized time series is generated based on the attribute input set (102) and a plurality of reference time series (108) associated with each of a reference set (106) of attributes for facilities different from the facility, The time series generator (100) includes a time series modification module (204) configured to generate the combined time series (104) by modifying the plurality of reference time series, wherein the reference set of attributes is determined to be similar to the input set of attributes, The aforementioned time series correction module (204) i) The system is configured to modify the multiple reference time series by calculating weighted combinations of the multiple reference time series and outputting the weighted combinations as the combined time series (104), where the weights of the weighted combinations represent the similarity between the input set of attributes and the reference set of attributes, or ii) A system configured to modify a plurality of reference time series by concatenating intervals selected from the plurality of reference time series, wherein the intervals are selected according to the similarity between a reference set of attributes associated with each of the plurality of reference time series and an input set of attributes.

2. The system according to claim 1, wherein the time series generator (100) comprises a reference data acquisition module (202) configured to search one or more databases to retrieve a plurality of candidate time series associated with each of the candidate attribute sets, and to select a plurality of candidate time series associated with attribute candidate sets similar to the input set of attributes (102) as the plurality of reference time series (108).

3. The system according to claim 2, wherein the reference data acquisition module (202) is configured to select a plurality of candidate time series as the plurality of reference time series (108) that, when compared with the input set of attributes (102), yield a similarity metric that satisfies a predetermined similarity threshold.

4. The system according to claim 3, wherein the reference data acquisition module (202) is configured to determine the similarity metric for each candidate set of attributes by calculating a distance metric representing the distance between the candidate set of attributes and the input set of attributes (102).

5. The system according to claim 3, wherein the reference data acquisition module (202) is configured to determine the similarity metric for each candidate set of attributes by inputting two such sets of attributes as feature vectors to a trained classifier trained to predict a similarity metric target based on an input feature vector comprising a candidate set of attributes and an input set (102) of attributes.

6. The system according to claim 1, wherein the time series correction module (204) is further configured to use an optimization algorithm for finding the optimal weights for the weighted combination.

7. The system according to claim 1, wherein the time series generator (100) is configured to generate the synthesized time series (104) by inputting the input set of attributes (102) into the machine learning model (304), which has been trained to predict the plurality of reference time series (108) based on each of the reference set of attributes (106) input into the machine learning model (304).

8. A system for synthesizing energy time-series data for a facility, wherein the system is The system includes a time series generator (100) configured to receive an attribute input set (102) having attributes that characterize the facility, and to output a synthesized time series (104) representing estimated energy data for the facility, wherein the synthesized time series is generated based on the attribute input set (102) and a plurality of reference time series (108) associated with each of a reference set (106) of attributes for facilities different from the facility, The time series generator (100) is further configured to modify the synthesized time series (104) in a post-processing step to match one or more predetermined key attributes of the facility, in a system.

9. The system according to any one of claims 1 to 8, wherein the attributes include one or more of the following: the classification of the facility, the annual energy consumption or generation at the facility, the operating hours, the geographical location, and the equipment used by the facility.

10. A method for synthesizing energy time series data for a facility, wherein the method involves a time series generator (100), The system receives an input set (102) of attributes that characterize the aforementioned facility, The system includes outputting a composite time series (104) representing estimated energy data for the aforementioned facility, and causing the system to perform the following: The synthesized time series is generated based on the input set of attributes (102) and a plurality of reference time series (108) associated with a reference set of attributes (106) of facilities different from the facility. The time series generator (100) includes a time series modification module (204) configured to generate the combined time series (104) by modifying the plurality of reference time series, wherein the reference set of attributes is determined to be similar to the input set of attributes, The aforementioned time series correction module (204) i) The system is configured to modify the multiple reference time series by calculating weighted combinations of the multiple reference time series and outputting the weighted combinations as the combined time series (104), where the weights of the weighted combinations represent the similarity between the input set of attributes and the reference set of attributes, or ii) A method configured to modify a plurality of reference time series by concatenating intervals selected from the plurality of reference time series, wherein the intervals are selected according to the similarity between a reference set of attributes associated with each of the plurality of reference time series and an input set of attributes.

11. A method for synthesizing energy time series data for a facility, wherein the method involves a time series generator (100), The system receives an input set (102) of attributes that characterize the aforementioned facility, The system includes outputting a composite time series (104) representing estimated energy data for the aforementioned facility, and causing the system to perform the following: The synthesized time series is generated based on the input set of attributes (102) and a plurality of reference time series (108) associated with a reference set of attributes (106) of facilities different from the facility. The method further comprises a time series generator (100) configured to modify the synthesized time series (104) in a post-processing step to match one or more predetermined key attributes of the facility.

12. A computer program, which, when executed by a computer, comprises instructions that enable the computer to carry out the method according to claim 10 or 11.

13. A computer-readable storage medium comprising instructions, when executed by a computer, that enable the computer to carry out the method according to claim 10 or 11.

Images